Pandemic AI Response (PAIR) Community

The American College of Radiology (ACR) Data Science Institute® (DSI) has developed a Pandemic AI Response (PAIR) community to leverage ACR and other resources in the near term battle against COVID-19. In the long term this effort will serve to develop a framework to prepare our clinical radiologists and researchers to combat potential future infectious disease outbreaks.

The radiology community has mobilized in response to the COVID-19 outbreak, and the new PAIR community is intended to foster and disseminate this knowledge going forward. The ACR is committed to working with all relevant imaging societies, other medical organizations, federal agencies and industry stakeholders to maximize our collective response.

Specific goals of the PAIR community are to provide the following resources:

Case Repository

The creation of a real-time COVID-19 case repository for immediate submission and distribution of COVID-19 cases between facilities throughout the United States. The COVID-19 case repository is freely available to all

ACR Assist Module for COVID-19 Reporting

The creation of structured reporting modules for COVID-19 to establish consistent reporting of radiographic findings of COVID-19 and aid radiologists in their communication with other healthcare providers assisting in the management of patients. An initial example was based on the RSNA Consensus Statement on Reporting Chest CT Findings Related to COVID-19, endorsed by ACR, and is available now.

AI Algorithm Repository and Local Algorithm Evaluation

Organizations throughout the world are creating AI algorithms in an attempt to solve many of the COVID-19 challenges. While nearly all are free to use, the ability to evaluate these solutions locally has been a barrier to most. Through ACR AI-LAB™ we will provide free hosting of any relevant COVID-19 AI algorithm for immediate evaluation by any facility in the United States. The growing list of algorithms can be found in AI-LAB. In addition to using AI-LAB resources to evaluate these algorithms against their own data, institutions will be able to provide feedback regarding the efficacy of the algorithms in their populations of patients.

Vendors or researchers interested in submitting COVID-19 related AI algorithms, please contact us at dsi@acr.org.

Federated Framework for Model Creation

We strive to enable multiple sites to collaborate on the creation of federated COVID-19 models, providing the power of large datasets without the need to move imaging data off-premises. To learn more about the ACR federated learning framework visit ACR AI-LAB 

Participation in National AI Registry

Case repositories are just the beginning. In conjunction with the ACR Center for Research and Innovation™  (CRI), the ACR will leverage our ability to transfer images and data from multiple sites for research and quality registries to create a multi-purpose, AI-relevant COVID-19 image and metadata national registry.

AI Use Case Development

A number of potential uses for AI in medical imaging have emerged including identification of imaging findings of COVID-19 and differentiation from other disease entities, quantification of disease burden and in potential triage for patients with clinically significant disease. The ACR DSI Thoracic Radiology Panel is developing structured AI use cases for a number of clinical, research and non-interpretive scenarios where AI could impact clinical care of COVID-19 patients. The panel is working closely with the ACR and RSNA RadElements.org team on the common data elements (CDEs) necessary for AI development and structured reporting. If you are interested in providing input to the panel or submitting an AI use case related to the COVID-19 pandemic, do so at ACR AI-LAB™.